Encrypting block driver for Linux...need some advice

[Johnathan Corgan <jcorgan@aeinet.com>]

-----BEGIN PGP SIGNED MESSAGE-----

All,

I dropped off the face of the earth for a few months while fighting a 
particularly *nasty* divorce, and to nurse my wounds, I immersed myself 
in writing cypto-code :)

What I've come up with is a loadable module block device driver for Linux 
that implements transparent encryption/decryption (is there a generic 
word that means both, like 'cryption' or some such?).

The way it works is by 'attaching' a filespec to the block driver, and 
translating block requests into read/write requests at the appropriate 
offset in the file.  During the read or write, the data is transformed 
with either DES or 3DES (RSAREF implementation).  The key is an MD5 hashed 
passphrase entered by the user when attaching the filespec.  The key is 
not stored anywhere, and there is no hidden structure to the ciphertext 
(such as a header block.)

The filespec can represent pretty much anything--another block device 
such as a hard drive partition or floppy drive, a regular file, a remote 
NFS exported file, CDROM, whatever.  If the file is remote, only ciphertext
is transmitted on the wire.

This part is working rather well at this point (as long as everything is 
done as root), which brings me to the crux of my post.

Being a Unix programming novice (lots of C experience on DOS/Windows), 
I'm pretty clueless when it comes to Unix level security issues.  I need 
to define and implement the appropriate behavior of the device when 
dealing with access to the data by different users.  Ideally, I want 
something infinitely flexible and configurable--why program in policy?--so 
that the user/admin can deal with a variety of threat models.

Another, more crypto related question--how to deal with IV's?  Right now, 
I'm using 512 byte sectors with CBC.  For each sector, the IV is the 
sector number.  This frustrates the known plaintext attack issue, but I'm 
not sure if such a simple scheme is really effective.  Probably not.

Then there is a whole host of issues relating to cryptographically 
hygienic programming practices...of which I am also pretty ignorant 
(especially on Unix.)

I guess you could say the software is at the "proof of concept" stage, 
and lacks all sorts of desirable features.  But it works (with many 
bugs I'm sure)...and I have to give credit to the Linux effort:  so far I've 
done this with nothing but the kernel source and the kernel hackers guide 
as a reference.  I took a look at doing this with Windows 95 and didn't 
even know where to start. (No, I'm not bashing Windows--I love Win95, use 
it all day at the office and get loads of work done with it--but Linux 
kernel hacking is much more fun.  An ideal world would have the Win95
UI/Plug & Play stuff coded on top of a Unix kernel :) 

In any case, suggestions, criticism, and comments are welcome.  The software
will eventually be GPL licensed when it is a bit more mature.

==
Johnathan Corgan       "For the first time in history, it is possible to
jcorgan@aeinet.com      have absolute privacy over arbitrary distances."
PGP Key Fingerprint:    4F 28 69 B8 76 2E 42 3E  8B 4C 12 BB 3A 43 D4 07


-----BEGIN PGP SIGNATURE-----
Version: 2.6.2

iQCVAwUBMBaCnU1Diok8GKihAQGyoQP+JAYyukotfejK84bm8olDs1GMd6zlwuXc
S+91DwrPRb8pyciEC6lIoLNS3coMgPdGTEksNNJMbuIXupJNnXnSum9XrPkMzEkG
gL/x6n6v4Jzm9B9IyvIV2R1UrIK893EGQbPKTIgGNNsvORJ/NB8nkoMfZalVlNnD
Hl3z3vaYgtU=
=grpJ
-----END PGP SIGNATURE-----